Marco Bertini, marketing professor at Esade Business School, says more and more companies are turning to pricing algorithms to maximize profits. But many are unaware of a big downside. The constant price shifts can hurt the perception of the brand and its products. He warns that overreliance on artificial intelligence and machine learning without considering human psychology can cause serious damage to the customer relationship. And he outlines steps managers should take, including implementing guardrails, overrides, and better communication tactics. With London Business School professor Oded Koenigsberg, Bertini wrote the HBR article “The Pitfalls of Pricing Algorithms.”
CURT NICKISCH: Welcome to the HBR IdeaCast from Harvard Business Review. I’m Curt Nickisch.
When I was a teenager on a sports team, our team bus stopped at a McDonald’s restaurant. Waiting in line, we noticed that the board that showed the prices for the French fries and hamburgers had little wheels for the numbers. Whoever set the prices, spun the wheels to show the right number. One of us joked that they probably rolled the prices higher when they saw the bus pulling up.
And as you do, when you’re trying to be funnier than the next kid, we started imagining the prices bouncing around like stocks that if you were shopping at McDonald’s, you’d have to yell, “Buy now!”, when the price of a big Mac dropped to where you wanted to swoop in and pick one up.
What was totally laughable to us at the time is actually become the norm of pricing today. Algorithms that can change prices by the minute are commonplace. It’s not just airfares and hotel rates anymore. Check your ride sharing app a minute later, and you can get a wildly different number. Leave items in your online shopping cart, and the next day you’ll find a new price. Now there’s a clear incentive for this. Companies eke more profit out of every transaction, but what many of them fail to understand is how much this dynamic pricing is messing with the psychology and trust of their customers.
Joining us today to explain the harm that dynamic pricing causes and how to manage it is Marco Bertini. He’s a marketing professor at ESADE Business School in Barcelona and a visiting professor at Harvard Business School. With Oded Koenigsberg of London Business School, Bertini also wrote the HBR article, “The Pitfalls of Pricing Algorithms”. Marco, thanks for coming on the show.
MARCO BERTINI: Oh, it’s a great pleasure to be here.
CURT NICKISCH: Marco, I have to admit, preparing for this interview, I just hadn’t realized just how fundamental pricing has changed since you know, I was younger.
MARCO BERTINI: Well, it’s changed. It’s changed a lot. I think in great part by the way, companies start to understand and actually individuals as well, but the way companies start to understand the many of the ramifications that a price has when you transact your customers. Back in the days when you were at school, it was fundamentally an economics principle, right? A bar that you set higher and lower and more people come in and more people are left out, but there is just much more knowledge and awareness and conscientiousness of the psychology behind numbers, which makes it a lot more richer. On top of that is the issue that technology is allowing us to do many, many more things that we could do before and we’d price. It was just kind of fun.
CURT NICKISCH: Yeah. I mean, price has always changed, right? And people waited for sales or they used coupons, but it just felt much more stable and certain and simple, maybe is the way to put it. We’re used to kind of wait to try to gain the best airfare. Now you can almost do it for, for anything. It’s not just what you’re going to buy and whether the price is worth it, but, but when you are going to buy it or at what price?
MARCO BERTINI: Some would say even maybe too much, right? But I think it’s good to have a little bit of context to these, because again, if our recent history is one of fixed prices, essentially, where a product has its price and that’s it. And you see it on a price tag. And that’s literally it.
But if you wind the clock back a century or so, it was with bazaars and marketplaces, there was no such thing as a fixed price. There was haggling, right? There was bargaining and every interaction between a customer or a firm led to some sort of price, depending on how price conscious that particular customer was and how much the business wanted to sort of extract that profit. Then what happened? A scale happened in a sense that we wanted to start selling lots more stuff, to lots more people at scale.
And still we started building stores. And when you started building stores and especially department stores, it’s really hard to bargain and barter and haggle inside a department store. It just becomes very, very messy. So to achieve that sort of scale, you start saying, oh, you know what, I’m going to put one price on this piece of item. So the customer doesn’t need to interact with me, the salesperson, and everything will be great. There’ll be less friction in the transaction.
So that’s kind of where we went. And that was stage two, let’s say, and then stage three is back to what I was alluding before. I can still with technology. I can still achieve that. This is where algorithms kind of come into it. I can still achieve that scale, that I so much desire, but I don’t necessarily have to stick to a price tag. And so I can, I can actually have the benefits of both worlds. And so in many senses, it’s kind of going back to the past, through the use of, of technology.
CURT NICKISCH: That’s amazing. I mean, there is a beauty, right, to the bazaar, price setting at a bazaar where, where the seller and the buyer haggle and basically the price is what the buyer and the seller agree it is at one moment in time. But if you take most consumers today, it seems a little exhausting or it makes, makes it feel much more adversarial somehow when you’re, when you’re trying to decide when to buy something.
MARCO BERTINI: Yeah, I would, I would definitely agree. And I think there’s a few things that are worth mentioning. One of them is it’s like a moving target a bit in the sense that social norms change all the time, right? If we wind the clock back aa sufficient amount of years, the thought of all of us paying different prices at a, at a concert, in an airplane, at a hotel was probably a foreign concept. And we would maybe take an objection to that.
But of course, that’s kind of sacred nature right now. We find that less intrusive, right? At the end of the day, commerce is a social phenomenon. I think, however, what you said is particularly true. There are situations where a company can take it too far. There are situations where it doesn’t have sort of the agency to do so, and the customer doesn’t really allow them implicitly to, to change prices so much or out of situations where customers actually look to the price for some sort of information. And when that information just keeps changing all the time, because the price keeps on changing, I’m left to fill in the blanks. And I would say, generally speaking, you do not want customers to fill in the blanks because often that leads down sort of a bad path.
CURT NICKISCH: Right, yeah. You get to that point where you think, oh, the company just noticed that I’m coming back to buy this and they raised the price.
MARCO BERTINI: Right. Exactly, exactly. But so often what I try to explain both to the students and to the companies, that I have the pleasure to work with is that when we think about these idea of customer focused, that it’s everybody knows about and everybody tries to strive for, if you’re a business, has two sides to it. It has like a front end and a back end to it.
So the front end of customer orientation is what we learn in marketing courses, which is: I’ve got a product or a service, and if I really want to do well in the marketplace, what I should be doing is understanding what the customer’s needs and wants and desires are, and then work my way backwards in terms of how to shape that product, how to communicate that product. So I’m being driven by the customer.
And so more and more, and again, especially with technology, organizations are thinking to themselves, well, how do I leverage technology to build that stronger relationship with customer and a stronger connection, which is great. In my opinion, there’s absolutely nothing wrong with that. The problem is that commerce is unique in the sense that at that moment, I then have to turn around and ask my customer for money. You know that saying, you know friends and money don’t mix everybody’s set up and everybody believes completely. In commerce, we don’t have the luxury of choosing one or the other. We kind of have to do both, right?
And so on the one hand, on the front end, as I was saying before, you’re trying to build that trust with customers, which is again, it’s great, but then it’s becomes very delicate, right? Because I’m telling my customer who I had previously told him or her to trust me a lot to then say, okay, now I used to give me money.
So it becomes a very, very difficult, very delicate thing to do. And if I start using algorithms to generate those prices, so prices are not only moving than they used to be, but also they’re moving in a way that may be disconnected. We had done the line psychology of that relationship. What can happen is that you’re driving a wedge between the front end of my customer orientation efforts and the back end, which is the more monetization element.
And whenever you drive that wedge going back to a comment I made before you start having the customer filling in the blanks, what’s going on? Why did they tell me these in the one hand, but then actually the pricing me that way? Are they standing behind what they’re saying? Are they truthful? Are they giving me what they promised?
CURT NICKISCH: Where does this go wrong sometimes? In your article, you gave the example of Uber at the time of a terrorist attack or the threat of a terrorist attack. Search pricing took the price of getting writers out of a certain area up five times what it normally is. There is some benefit of course, to offering, charging more money and drawing more drivers to help take people out of a certain area and at a high demand shock event like that, but what went wrong in that case, do you think?
MARCO BERTINI: So, I think if we, put yourself in the customer’s shoes. In the case of Uber, when these shocks happen, sometimes even if it is not, even if it’s not a harmful sort of situation, but more like a Christmas Eve or things, or New Year’s Eve, there is that sense, I think, from the customer perspective that it’s not as if I’ve got many options, right. I am, I’m at the mercy of the company. I am out here. I need to get out of here for whatever reason, and I literally do not have a choice.
CURT NICKISCH: So what do you think Uber should do in a situation like that? Because it could still offer high reimbursement rates, right. To drivers to drive on New Year’s Eve, for instance, is that just one of those times when Uber should just lose money for the evening?
MARCO BERTINI: I mean that may be a solution, right? I don’t pretend to understand the fundamental economics of Uber, so I wouldn’t hesitate to sort of give them specific advice. But from where I sit, my objective would be to sit down with them. I mean, we make an example of Uber, but it could be this there’s millions of applications of algorithms. What I ideally like is for organizations to sit down and think beyond the standard economic argument of algorithms. That’s, because that, I think they’re in the best seat to judge what limits there are. If they’re made aware of those sort of side effects, or byproducts, I think they’re in the best seat to sit down and say, okay, oh, now that I know that these things could happen, and given what I know about the marketplace, where do I want to put those guardrails, for example. Or when should I expect something to potentially go wrong, and so to be weary of that? Or who should I put the decision in the hands of?
You see what I mean? It’s hard, right, to say I’m not Uber, so it’s hard for me to say, “Hey, oh, this is what he should have done.”. Yeah. Accept the loss for that day, because that’s the way you should go about it. I think I’m usually satisfied when you, in this specific case, anyway, when you get sort of an organization to think, okay, I didn’t know that when I set, when I vary my price, that releases information, beyond the fact by no buy now, buy later, wait, wait a second, by now. Beyond that sort of basic information, I wasn’t really aware that actually there is more information about that. And if we say internalize that idea, then, hey, all of a sudden you can have lots of different options. And importantly, some of those options may actually run against what you would do if you’re looking at it purely from an economic perspective, short term economics.
CURT NICKISCH: Let’s talk briefly about algorithms themselves, why they’re so powerful and so popular now, and is some of this evolution, just something that you see that can be built into algorithms going forward, that we just haven’t got there yet with how pricing is done?
MARCO BERTINI: Yeah. I really think that is the case. I mean, everything has its evolution, and the idea of using an algorithm to help me do a function that I was doing probably on a spreadsheet beforehand. It makes perfect sense back to what I was saying before. But again, the first intuition is I’m helping, I’m trying to get this algorithm to help me figure out where the brakes on the willingness to pay are, and sort of understand exactly what price fits better to what customer. That is the core, and it’s always going to be the core.
CURT NICKISCH: And that’s pretty powerful, right. I mean, that’s yeah.
MARCO BERTINI: Absolutely.
CURT NICKISCH: And just in today’s world with supply chains or maximizing the number of seats on a plane, it’s just, it’s crucial.
MARCO BERTINI: Absolutely. Right. Because either we have a fixed capacity that we would like to sort of make the most of or because, and I always tell this to my students, I want to move people around so that there’s a case just a second. There was a case of Disney in the article. And if I see myself with having seasonal demand, where very lots of people trying to come to my theme parks in the middle of summer or during the school vacations and less people going other times, I can use the algorithm, of course, to have a financial return by all means, but I’m actually using it to change the customer experience as well. Because if I can move people around and get families to come to realize there’s a trade-off between coming when you would really like to come, but maybe it’s the school vacation and another moment and accept that trade off.
Then what happens is that this is analogy so this appears a little bit, and everybody’s experience is actually much better because we didn’t have to wait so much in line. So I guess my point I’m trying to drive is that, there is a core element to algorithms, which are just fundamental economics 101 just used with just done a bit better with, or sometimes even a lot better through technology, but then there’s this periphery of psychology and sociology that all mixes with that, that it’s typically a periphery, but in many, many cases actually can come to dominate. You got these Uber-like individual events that are, you know, so outrageous that just, you know, they just come to dominate. And they sometimes unfortunately shape the perceptions of the company, even though maybe only happened once or a few times.
CURT NICKISCH: So what can be done to improve the machine learning that’s behind these algorithms to better incorporate, recognize human psychology? Is it, what is the answer here?
MARCO BERTINI: The answer, like many good answers tends to have multiple levels. The first thing is just, and I’m always very big on this, is just awareness of the problem, because if we don’t have awareness of that problem, that is like, there’s no reason, I’m just trying to push something down your throat that you don’t even understand why I’m pushing it down your throat.
So the first thing is to understand, in my opinion in my experience, most organizations do not think of prices beyond the sort of the mechanics of it and the numbers of it. Right? So there’s much more than a numbers thing. So, just like it’s more than a numbers thing when you’re setting prices yourself on an Excel and you change them every six months. It is also more than the mechanics when you’re changing it every second.
And guess what, it’s much more important when you’re changing it every second, because there’s a lot more stuff moving and lots of more inferences making. So we have to understand that first. Once we understand that first, then I get second, you want to say, well, okay. So there is inferences beyond the beyond, just when I should buy and how I should buy. Okay. Then what do I want to say? How do I make sure, from my perspective, I think it’s a two double, double-edged sword here or a two-way street maybe is the better analogy. The one way of the street is how do I use the algorithm to make sure that whatever my brand is, it resonates through the algorithm.
The other side of the street is understanding that the same thing happens the other way around. When I vary prices and my customers respond, that is like experiments. It’s a great experiment. Because people actually putting their money where their mouth are.
So, do I perceive, and this is a bit tricky. Do I use dynamic pricing of course, to get my products in the hands of customers, because that’s what I want primarily. But at the same time, I may actually proactively thinking about it as I, as an experiment where the variation in price is actually gives me a sense of what people are responding to in terms of their appreciation of a product.
Because if I bring something new to market and I really want to understand what they really care about, this one, a great way of learning about it is through some variation in prices, and then seeing how behaviors respond.
CURT NICKISCH: There can be a lot of benefits to dynamic pricing to the consumer. How should companies go about showing the benefits of that or making the decision making that they’re doing, in some cases on behalf of the consumer or to the consumer’s benefit, more apparent to the people they’re reaching.
MARCO BERTINI: Right. So that’s a great question. And it kind of strikes at a sort of a bigger, I would call it problem. Whenever a company tries to move away from one price to all of its customers, to a series of prices, to its customers. And it does that because it realizes that people just have different valuations for the things that you sell. Some people like it more, some people like it less, that’s just nature. So whenever we start thinking about those kinds of things, I think many of us, maybe most of us, jump up and say, oh, see, that’s just very nasty, right? The company’s trying to exploit. And the fact that I have a high willingness to pay and just, it’s just not nice.
I mean, I’m biased because I’m a pricing person, but I think it gets a bit of a bad rep, right. Because what happens is that I’m basically de-averaging something. So I’m the de-averaging the price. And yes, it’s true that some people, if I am able to, will end up paying more for something that we’re paying less for beforehand, but hold on a second, there’s actually people that are paying, are now asked to pay less for the same product or service. And maybe in beforehand, they were not able to afford accessing the part of the service, but now they do.
Think of categories where access becomes, it’s of basic importance. Maybe healthcare, maybe education, maybe insurance. I can just sort of line up a few of these, right? So in situations where you really want to get all the market covered, because they really should have coverage. Pretty much, the only way you can do is either A, I bring the price, the one price down to zero, in which case you’ve got this dilemma that I’m actually not making any money, and therefore, as a company I might not be alive very long.
Or I flex my prices as much as I can such that there is some sort of cross subsidization going on in the marketplace. And you had the ones that are willing to pay more able to pay more. So cox subsidize the ones that are willing or able to pay less. So I guess what I’m getting at is that one huge advantage of flexing your prices, which can then be done dynamically through algorithms, is that it just broadened access so much more, and in some categories, that is really what you’re after.
CURT NICKISCH: Yeah. That’s interesting. And then you have to communicate that somehow.
MARCO BERTINI: Exactly. So that’s the bad rep part is that maybe what you haven’t done very well is explain it. And I’m thinking of an example here as just as I’m talking, I know these at least one business in the providing food. So I’d say as a chain of restaurants providing healthy food, and part of its mission is to make sure that even households with a lower income who may otherwise go to a fast food place and the fast food perhaps is not the best type of food to eat on a regular basis, even lower income households have access to more nutritional meals. And so they’re very open about the fact that what they’ll do is they will charge for the very same bowl, for the very same salad, for the very same main course, whatever it may be. For the very same plate, they will charge a very different price within different areas of the city, in which there might be in, trying to judge, depending on the area, the zip codes, what the purchasing power may be of that particular neighborhood, right. And so they’re very upfront about the fact that we’re going to charge different prices because our goal is to have nutritional food available to all. And that is kind of a higher goal then there might be sort of just keeping those prices constant and maintain a certain profitability.
CURT NICKISCH: When you’re talking about this high level now, I understand those, I understand that perspective kind of as a leader in an organization how you’re trying to think about that. What are some practical things that you can do then to follow through on this?
MARCO BERTINI: Yeah, absolutely. So that was like the step two. I think we mentioned before, I says they step three and beyond there are the more we’re slowly landing this other, the step three and more and more of the practical, the practical aspects. So, and some of those things came out in our discussion already. So the idea of guardrails, I think it’s kind of established now and it’s an important thing to sort of stress.
But some understanding of where the variation may arrive to a point that is just it’s extreme, and, therefore, the customer will start saying, what is going on here? Remember, one of the things that I was stressing is that you don’t necessarily want your customers to fill in the blanks. So if A, if you’ve got a narrative that you can explain why these variations are there, then by all means, but if there is no such thing as a narrative, then you want to make sure that you sort of you’ll hold yourself to certain patterns.
Related to that one is the notion that you really want your algorithm to be connected in a much stronger way than it probably is right now to the strategic decisions of the organization. And maybe even better said, the marketing of positioning value proposition ideas of the organization. And what do I mean by that?
So generally speaking, and maybe again, maybe we’re generalizing a bit too much, but I think it’s fair. Generally what you tend to have is that these algorithms are, it’s just a software that I’ve heard from some people that I should be applying in my business. So I do some research, of course, and then I bring it in and most likely have a data team that is managing it and helping me with that, maybe some of the pricing folks as well. And that’s great, nothing wrong, absolutely wrong with that. But that sort of structure is the one that will maximize the economic aspect of these algorithms. The more psychological slash sociological aspects that relate to inferences about the company and the products, those are not typically the expertise doesn’t lie in those particular functions of the business.
CURT NICKISCH: You know, this has been really helpful, Marco. Can you give one example, just shining example of a company that you saw do, something that you thought was really smart when it came to trying to solve this problem?
MARCO BERTINI: I always liked the way Disney at least overtly was approaching the topic, because it thinks about the topic in an even more strategic way. Of course, there is the element of like, we’ve been saying across this interview that the more I’m able to flex my prices properly, given the right kind of variables, then I can sort of really customize that price. And that is great. But then there is the customer experience angle, which we had mentioned whereby what’s actually happening is that if I provide folks with a menu of different prices that are moving dynamically, then people will make a decision when to attend.
And what I can do is I can smooth out demand, hopefully. And then what is interesting to me, maybe as I came more as an academic is I am using my prices to change the experience of the customer. That is I’m using my prices to change the value they derive from the product itself, which is weird because it should be that way around. You have a volume, then you capture through your price, but actually the way I price actually changes the value itself in a sense. And then on top of that, there’s also a third element. So there is experience element, there is the sort of revenue element, and then there is a cost element, right? Because if I change, again, the behavior of customers, if I’m successful at that, through my use of my algorithm, I’m also lowering my cost base.
And so I kind of have this sort of triple whammy effect and, it kind of strikes at the heart of most of the things we try to do in business in general but in particular when it comes to pricing, which is whatever dollar I invest in something think about discount or whatever dollar of discount that I invest, I want to get the biggest return possible. And so the more things I can get out of that $1 of investment, and in this case, the more I can get out of the algorithm doing its thing, the better it is. And they’re pretty smart about it in my opinion.
CURT NICKISCH: Marco, this has been fantastic. Thanks for sharing these insights with us.
MARCO BERTINI: Thank you very much for listening to me.
CURT NICKISCH: That’s Marco Bertini. He’s a marketing professor ESADE business school, and he’s a coauthor of the article, “The Pitfalls of Pricing Algorithms”. You can find it in the September/October 2021 issue of Harvard Business Review and at HBR.org.
And to hear more on pricing, check out and to hear more on this topic, check out pricing strategies for uncertain times, that’s episode 748.
This episode was produced by Mary Dooe. We get technical help from Rob Eckhardt. Adam Buchholz is our audio manager. Thanks for listening to the HBR IdeaCast. I’m Curt Nickisch.